CN106056648A - Intelligent robot image drawing method and system - Google Patents

Abstract

The invention discloses an intelligent robot image drawing method and system. The method comprises the following steps: inputting a source image to be drawn; carrying out feature extraction on the source image through a mixed binaryzation image processing method to obtain a region to be filled of a robot; carrying out region segmentation and optimal path planning through a path optimization algorithm according to the region to be filled of the robot; and carrying out robot image drawing according to the obtained optimal drawing path. The source image is subjected to feature extraction through the mixed method based on local binarization and global binarization, and the method is high in robustness and reliable and accurate; and with the shortest total drawing time being a target, the region segmentation and optimal path planning are carried out through the improved path optimization algorithm, and thus the optimal drawing path of the robot is obtained, and the drawing time is shorter. The method can be widely applied to the field of picture processing.

Description

The image drawing method of a kind of intelligent robot and system

Technical field

The present invention relates to image processing field, the image drawing method of a kind of intelligent robot and system.

Background technology

Current intelligent robot has become a popular research topic, is applied in many commercial Application, as assembled, Spray paint and polishing etc..Along with intelligent robot technology and the development of drawing technique, robot drawing technique is arisen at the historic moment.Machine Device people's drawing technique uses industrial robot to draw the image captured by photographing unit.Owing to needing according to the image shot machine Device people control in real time, and robot drawing technique can be continually changing the situation of environment with the reply of show robot intelligence.

In robot drawing technique, image characteristics extraction and determine that drawing path is two most important steps.? In the most numerous image characteristic extracting methods, binaryzation feature extraction is a kind of more common method, and binaryzation feature carries Follow the example of and include overall situation binarization method and local binarization method.Overall situation binarization method algorithm is simple, but to because of uneven illumination Etc. situation, the treatment effect of the image that difference in brightness is big is poor, and robustness is more weak；Local binarization method is not by illumination condition Etc. the impact of factor, but its performance in some aspects is similar with high pass filter, may result in what robot was finally drawn Image is made mistakes, less reliable and accurate, and such as, one has large-area human hair's image and using at local binarization method Robot may be made after reason to draw one " bareheaded ".

Obtaining feature after image characteristics extraction is the closure region needing to be filled by robot, and there is complexity in these regions Shape and cavity, directly process these regions and become extremely difficult, thus travel through within the shortest time region of these complexity with Determine that the drawing path of robot has just become a technical barrier needing solution badly in the industry.Existing robot drawing technique is to this Good solution, not can not meet people's high request to image rendering speed.

Summary of the invention

For the above-mentioned technical problem of solution, it is an object of the invention to: a kind of strong robustness, reliable, accurate and drafting are provided Time is short, the image drawing method of intelligent robot.

It is an object of the invention to: provide a kind of strong robustness, reliable, accurate and the drafting time is short, intelligent robot Image drawing system.

The technical solution used in the present invention is:

The image drawing method of a kind of intelligent robot, comprises the following steps:

Input source images to be drawn；

Use based on mixing binary image process method source images is carried out feature extraction, obtain robot wait fill out Fill region；

Region to be filled according to robot uses path optimization's algorithm to carry out region segmentation and optimum path planning, obtains The optimal drawing path of robot；

Robot carries out image rendering according to the optimal drawing path obtained.

Further, the method that described employing processes based on mixing binary image carries out feature extraction to source images, obtains Include the step for of the region to be filled of robot:

Use local binarization method that source images is processed, obtain local binarization image；

Use overall binaryzation method based on histogram thresholding that source images is processed, obtain overall situation binary image；

Local binarization image and overall situation binary image are merged, obtains including the region to be filled of robot Binary image；

Binary image after merging is vectogram by bitmap-converted.

Further, source images is processed by described employing local binarization method, obtains this step of local binarization image Suddenly, comprising:

Using integrogram method to calculate the gray scale arithmetic mean of instantaneous value of the neighborhood of each pixel in source images, computing formula is:

Wherein, (x is y) that ((x is y) that (x, y) at integrogram at source images midpoint to I for x, pixel value y) at source images midpoint to p The pixel value of middle correspondence, s be source images midpoint (x, the size of neighborhood y),For rounding downwards symbol, w and h is respectively For width and the height of source images, α and β is given weight parameter, and (x y) is source images midpoint (x, the ash of neighborhood y) to m Degree arithmetic mean of instantaneous value；

Each pixel that meansigma methods is source images according to calculating calculates corresponding local threshold, and computing formula is: T ((x, y)-c, wherein, (x is y) that (c is given constant for x, local threshold y) at source images midpoint to T for x, y)=m；

Local threshold according to calculating carries out local threshold to source images, obtains local binarization image, described source figure The local threshold formula of picture is:

Wherein, (x y) is source images midpoint (x, y) pixel value after local threshold to b.

Further, source images is processed by described employing overall binaryzation method based on histogram thresholding, obtains the overall situation The step for of binary image, comprising:

Obtain the histogram curve of source images；

Histogram curve according to source images uses jitter elimination method to determine the rectangular histogram global threshold of source images；

According to the global threshold determined, source images is carried out binaryzation, obtain overall situation binary image.

Further, the described histogram curve according to source images uses shake diagnostic method to determine the rectangular histogram overall situation of source images Include the step for of threshold value:

Find all extreme points of the histogram curve of source images, and set up the Candidate Set of valley point with this；

Judge in the Candidate Set of valley point, whether the gray value of each extreme point meets the jitter elimination condition set one by one, if It is then this extreme point to be removed from the Candidate Set of valley point, otherwise, then retain this extreme point, finally give the candidate of non-jitter Collection, the described jitter elimination condition set as: | f (x_u)-f(x_u+1) | ＜ ε, wherein, x_uAnd x_u+1It is respectively in the Candidate Set of valley point U extreme point and the u+1 extreme point, f (x_u) and f (x_u+1) it is respectively x_uAnd x_u+1Corresponding gray value, u=1,2 ..., p- 1；P be valley point Candidate Set in the sum of extreme point, ε is the jitter parameter set；

Find out first from the Candidate Set of non-jitter and meet x_oFor maximum point and x_o+1Extreme point x for minimum point_o, so After with minimum point x_o+1Corresponding gray value f (x_o+1) as the rectangular histogram global threshold of source images, wherein, o=1,2 ..., q- 1；Q be non-jitter Candidate Set in the sum of extreme point, and q≤p.

Further, the described region to be filled according to robot uses path optimization's algorithm to carry out region segmentation and optimum road Footpath is planned, the step for of obtaining the optimal drawing path of robot, comprising:

Use the edge fitting method improved that the edge in the region to be filled of robot is fitted, obtain treating after matching Filling region, the edge fitting method employing multi-section circular arc of described improvement carrys out every section of curve at the edge in matching region to be filled, and Use the line segment obtained by connection two end points of this section of circular arc to approximate this when the radius of circular arc is less than the radius threshold set Section circular arc；

It is multiple inside subregion without cavity by the area to be filled regional partition after matching；

With the shortest total drafting time as target, use heuristic search algorithm determine each sub regions drawing order and The drafting starting point of each sub regions and drafting terminal, thus obtain the optimal drawing path of robot.

Further, described with the shortest total drafting time as target, use heuristic search algorithm to determine each sub regions The drafting starting point of drawing order and each sub regions and drafting terminal, thus obtain this step of optimal drawing path of robot Suddenly, comprising:

Find the beeline between any two subregion；

Every sub regions is abstracted into a point, using the beeline between two sub regions as the weight on limit, will be the shortest Total drafting time is found problem and is converted into traveling salesman problem, then uses heuristic nearest neighbor algorithm to solve each sub regions Drawing order；

Calculating total drafting time in each pattern path in each sub regions, computing formula is:

$\left\{\begin{array}{c}{t}_{total}={\&Sigma;}_{r=1}^n{t}_r+(n-1)t_c\\ t_r=\left\{\begin{array}{c}2\sqrt{\frac{l_r}{a}},l_r<\frac{v^2}{a}\\ \frac{l_r}{v}+\frac{v}{a},l_r\&GreaterEqual;\frac{v^2}{a}\end{array}\right.\\ t_c=2\sqrt{\frac{width}{a}}\end{array}\right.,$

Wherein, t_totalFor always the drawing the time of present mode path in current sub-region, v be robot drawing instrument Big translational speed, a is the acceleration that robot drawing instrument is given, and n is present mode path base line in current sub-region Quantity, t_cMove to the time of next base line, t from a base line for robot drawing instrument_rAnd l_rIt is respectively current son Filling time of r base line and length in present mode path in region, width is present mode road in current sub-region The width of footpath base line；

The shifting between different subregions is calculated with the translational speed of robot drawing instrument according to the distance between different subregions The dynamic time；

In drawing order according to each sub regions, each sub regions, the total of the path of each pattern draws time and difference Traveling time between subregion is modeled and optimization, obtains painting of each sub regions corresponding to the shortest total drafting time Starting point processed and drafting terminal, the shortest described total drafting time T_totalRelational expression be:

$\left\{\begin{array}{c}{T}_{total}=\underset{i,j}{min}\{D\&lsqb;n^{\&prime;},i\&rsqb;+T(cp\&lsqb;n^{\&prime;},i\&rsqb;,cp\&lsqb;n^{\&prime;},j\&rsqb;)\}\\ st\&lsqb;n^{\&prime;}\&rsqb;=cp\&lsqb;n^{\&prime;},i\&rsqb;\\ e\&lsqb;n^{\&prime;}\&rsqb;=cp\&lsqb;n^{\&prime;},j\&rsqb;\\ D\&lsqb;k,i\&rsqb;=\underset{j,t}{\min }\{D\&lsqb;k-1,i\&rsqb;+T(cp\&lsqb;k-1,j\&rsqb;,cp\&lsqb;k-1,t\&rsqb;)+T(cp\&lsqb;k-1,t\&rsqb;,cp\&lsqb;k,i\&rsqb;)\}\\ st\&lsqb;k-1\&rsqb;=cp\&lsqb;k-1,j\&rsqb;\\ e\&lsqb;k-1\&rsqb;=cp\&lsqb;k-1,t\&rsqb;\\ D\&lsqb;0,i\&rsqb;=0\end{array}\right.,$

Wherein, every sub regions is according to the corresponding group of drawing order, and n ' rolls into a ball for last, and k is kth group, k=1, 2 ..., n '；D [n ', i], D [k, i] and D [k-1, i] are respectively from the drafting starting point of first group a to n ', group k and group k-1 The shortest drafting time of some i, cp [n ', i] and cp [n ', j] is respectively group corner point i and j, st [n '] of n ' and e [n '] and divides Not Wei the drafting starting point of group n ' and draw terminal, T (cp [n ', i], cp [n ', j]) is the some j in the some i a to n ' in group n ' Drafting time, cp [k-1, j] and cp [k-1, t] are respectively the corner point i, T that corner point j and t, cp [k, i] is group k of group k-1 (cp [k-1, j], cp [k-1, t]) is the drafting time of the some t in the some j a to k-1 in group k-1, T (cp [k-1, t], cp [k, I]) it is the traveling time putting i in the some t a to k in group k-1, D [0, i] is the initial value of D [k, i].

What the present invention was taked another solution is that

A kind of image drawing system of intelligent robot, including:

Input module, for inputting source images to be drawn；

Levy extraction module, for using the method processed based on mixing binary image that source images is carried out feature extraction, Obtain the region to be filled of robot；

Optimal drawing path acquisition module, uses path optimization's algorithm to carry out district for the region to be filled according to robot Regional partition and optimum path planning, obtain the optimal drawing path of robot；

Image rendering module, carries out image rendering for robot according to the optimal drawing path obtained.

Further, described optimal drawing path acquisition module includes:

Edge fitting unit, for using the edge fitting method of improvement to intend the edge in the region to be filled of robot Closing, obtain the region to be filled after matching, the edge fitting method of described improvement uses multi-section circular arc to come matching region to be filled Every section of curve at edge, and connection two end points gained of this section of circular arc are used when the radius of circular arc is less than the radius threshold set To line segment approximate this section of circular arc；

Cutting unit, being used for the area to be filled regional partition after matching is multiple inside subregion without cavity；

Optimal drawing path determines unit, for the shortest total drafting time as target, uses heuristic search algorithm true Determine the drawing order of each sub regions and the drafting starting point of each sub regions and draw terminal, thus obtaining the optimal of robot Drawing path.

Further, described optimal drawing path determines that unit includes:

Beeline finds subelement, for finding the beeline between any two subregion；

Drawing order solves subelement, for every sub regions being abstracted into a point, and the shortest with between two sub regions Searching of the shortest total drafting time problem as the weight on limit, is converted into traveling salesman problem by distance, then use heuristic recently Adjacent Algorithm for Solving goes out the drawing order of each sub regions；

Very first time computation subunit, total drafting time in each pattern path in calculating each sub regions, calculates Formula is:

$\left\{\begin{array}{c}{t}_{total}={\&Sigma;}_{r=1}^n{t}_r+(n-1)t_c\\ t_r=\left\{\begin{array}{c}2\sqrt{\frac{l_r}{a}},l_r<\frac{v^2}{a}\\ \frac{l_r}{v}+\frac{v}{a},l_r\&GreaterEqual;\frac{v^2}{a}\end{array}\right.\\ t_c=2\sqrt{\frac{width}{a}}\end{array}\right.,$

Wherein, t_totalFor always the drawing the time of present mode path in current sub-region, v be robot drawing instrument Big translational speed, a is the acceleration that robot drawing instrument is given, and n is present mode path base line in current sub-region Quantity, t_cMove to the time of next base line, t from a base line for robot drawing instrument_rAnd l_rIt is respectively current son Filling time of r base line and length in present mode path in region, width is present mode road in current sub-region The width of footpath base line；

Second Time Calculation subelement, for according to the distance between different subregions and the mobile speed of robot drawing instrument Degree calculates the traveling time between different subregion；

Draw starting point and drafting terminal determines unit, for according in the drawing order of each sub regions, each sub regions Traveling time between total drafting time in the path of each pattern and different subregion is modeled and optimization, obtains The drafting starting point of each sub regions that short total drafting time is corresponding and drafting terminal, the shortest described total drafting time T_totalRelation Expression formula is:

$\left\{\begin{array}{c}{T}_{total}=\underset{i,j}{min}\{D\&lsqb;n^{\&prime;},i\&rsqb;+T(cp\&lsqb;n^{\&prime;},i\&rsqb;,cp\&lsqb;n^{\&prime;},j\&rsqb;)\}\\ st\&lsqb;n^{\&prime;}\&rsqb;=cp\&lsqb;n^{\&prime;},i\&rsqb;\\ e\&lsqb;n^{\&prime;}\&rsqb;=cp\&lsqb;n^{\&prime;},j\&rsqb;\\ D\&lsqb;k,i\&rsqb;=\underset{j,t}{\min }\{D\&lsqb;k-1,i\&rsqb;+T(cp\&lsqb;k-1,j\&rsqb;,cp\&lsqb;k-1,t\&rsqb;)+T(cp\&lsqb;k-1,t\&rsqb;,cp\&lsqb;k,i\&rsqb;)\}\\ st\&lsqb;k-1\&rsqb;=cp\&lsqb;k-1,j\&rsqb;\\ e\&lsqb;k-1\&rsqb;=cp\&lsqb;k-1,t\&rsqb;\\ D\&lsqb;0,i\&rsqb;=0\end{array}\right.,$

Wherein, every sub regions is according to the corresponding group of drawing order, and n ' rolls into a ball for last, and k is kth group, k=1, 2 ..., n '；D [n ', i], D [k, i] and D [k-1, i] are respectively from the drafting starting point of first group a to n ', group k and group k-1 The shortest drafting time of some i, cp [n ', i] and cp [n ', j] is respectively group corner point i and j, st [n '] of n ' and e [n '] and divides Not Wei the drafting starting point of group n ' and draw terminal, T (cp [n ', i], cp [n ', j]) is the some j in the some i a to n ' in group n ' Drafting time, cp [k-1, j] and cp [k-1, t] are respectively the corner point i, T that corner point j and t, cp [k, i] is group k of group k-1 (cp [k-1, j], cp [k-1, t]) is the drafting time of the some t in the some j a to k-1 in group k-1, T (cp [k-1, t], cp [k, I]) it is the traveling time putting i in the some t a to k in group k-1, D [0, i] is the initial value of D [k, i].

The beneficial effects of the method for the present invention is: use the method processed based on mixing binary image to carry out source images Feature extraction, has merged local binarization method and the advantage of overall situation binaryzation method, has solved overall situation binaryzation by local binarization The problem that the treatment effect of the image that method is big to the difference in brightness because of situations such as uneven illuminations is poor, not by the shadow of illumination condition Ringing, robustness is good；Local binarization method is avoided to bring because behaving like high pass filter by overall situation binaryzation local Problem, reduces the image finally drawn of robot and makes mistakes probability, relatively reliable and accurate；Region to be filled according to robot is adopted Carry out region segmentation and optimum path planning with path optimization's algorithm, obtain the optimal drawing path of robot, it is proposed that for Robot closure is filled path optimization's algorithm in region and is filled with, and can travel through the region of these complexity within the shortest time To determine the optimal drawing path of robot, draw the time shorter.Further, use the edge fitting method improved to robot The edge in region to be filled is fitted, and can be carried out every section of curve at the edge in matching region to be filled by multi-section circular arc, reduces The phenomenon of edge sawtooth, relatively reliable.Further, use heuristic search algorithm determine each sub regions drawing order and The drafting starting point of each sub regions and drafting terminal, thus obtain the optimal drawing path of robot, further by heuristic Searching algorithm improves traversal speed and reduces computational complexity.

The system of the present invention provides the benefit that: use the side processed based on mixing binary image in levying extraction module Method carries out feature extraction to source images, has merged local binarization method and the advantage of overall situation binaryzation method, has passed through local binarization Solve the problem that the treatment effect of the overall situation binarization method image big to the difference in brightness because of situations such as uneven illuminations is poor, no Being affected by illumination condition, robustness is good；Avoid local binarization method because behaving like high pass by overall situation binaryzation local Wave filter and the problem brought, reduce the image that robot finally draws and make mistakes probability, relatively reliable and accurate；Draw optimal In the acquisition module of path, the region to be filled according to robot uses path optimization's algorithm to carry out region segmentation and optimal path rule Draw, obtain the optimal drawing path of robot, it is proposed that the path optimization's algorithm filling region for robot closure enters Row is filled, and can travel through the region of these complexity to determine the optimal drawing path of robot within the shortest time, draws the time more Short.Further, the edge in the region to be filled of robot is carried out by the edge fitting method using improvement in edge fitting unit Matching, can carry out every section of curve at the edge in matching region to be filled by multi-section circular arc, reduces the phenomenon of edge sawtooth, more may be used Lean on.Further, determine that in unit, employing heuristic search algorithm determines the drawing order of each sub regions at optimal drawing path And the drafting starting point of each sub regions and draw terminal, thus obtain the optimal drawing path of robot, further by opening Hairdo searching algorithm improves traversal speed and reduces computational complexity.

Accompanying drawing explanation

Fig. 1 is the overall flow figure of the image drawing method of a kind of intelligent robot of the present invention；

Fig. 2 is the general frame figure that phase system is drawn by embodiment one robot；

Fig. 3 is the flow chart of embodiment one feature extraction；

Fig. 4 is the rectangular histogram of embodiment one source images；

Fig. 5 is the relation schematic diagram of embodiment one drawing path；

Fig. 6 is the different path mode schematic diagrams of the single subregion of embodiment one；

Fig. 7 is the experimental system structural representation of embodiment two；

Fig. 8 is the flow chart of steps of embodiment two feature extraction；

Fig. 9 is the process schematic that embodiment two robot carries out path planning and drawing；

The image that Figure 10 is drawn by embodiment two robot.

Detailed description of the invention

With reference to Fig. 1, the image drawing method of a kind of intelligent robot, comprise the following steps:

Input source images to be drawn；

Use based on mixing binary image process method source images is carried out feature extraction, obtain robot wait fill out Fill region；

Region to be filled according to robot uses path optimization's algorithm to carry out region segmentation and optimum path planning, obtains The optimal drawing path of robot；

Robot carries out image rendering according to the optimal drawing path obtained.

Being further used as preferred embodiment, described employing method based on mixing binary image process is to source images Carry out feature extraction, obtain including the step for of the region to be filled of robot:

Use local binarization method that source images is processed, obtain local binarization image；

Use overall binaryzation method based on histogram thresholding that source images is processed, obtain overall situation binary image；

Local binarization image and overall situation binary image are merged, obtains including the region to be filled of robot Binary image；

Binary image after merging is vectogram by bitmap-converted.

Being further used as preferred embodiment, source images is processed by described employing local binarization method, obtains office The step for of portion's binary image, comprising:

Using integrogram method to calculate the gray scale arithmetic mean of instantaneous value of the neighborhood of each pixel in source images, computing formula is:

Wherein, (x is y) that ((x is y) that (x, y) at integrogram at source images midpoint to I for x, pixel value y) at source images midpoint to p The pixel value of middle correspondence, s be source images midpoint (x, the size of neighborhood y),For rounding downwards symbol, w and h is respectively For width and the height of source images, α and β is given weight parameter, and (x y) is source images midpoint (x, the ash of neighborhood y) to m Degree arithmetic mean of instantaneous value；

Each pixel that meansigma methods is source images according to calculating calculates corresponding local threshold, and computing formula is: T ((x, y)-c, wherein, (x is y) that (c is given constant for x, local threshold y) at source images midpoint to T for x, y)=m；

Local threshold according to calculating carries out local threshold to source images, obtains local binarization image, described source figure The local threshold formula of picture is:

Wherein, (x y) is source images midpoint (x, y) pixel value after local threshold to b.

Being further used as preferred embodiment, described employing overall binaryzation method based on histogram thresholding is to source images Process, the step for of obtaining overall situation binary image, comprising:

Obtain the histogram curve of source images；

Histogram curve according to source images uses jitter elimination method to determine the rectangular histogram global threshold of source images；

According to the global threshold determined, source images is carried out binaryzation, obtain overall situation binary image.

Being further used as preferred embodiment, the described histogram curve according to source images uses shake diagnostic method to determine Include the step for of the rectangular histogram global threshold of source images:

Find all extreme points of the histogram curve of source images, and set up the Candidate Set of valley point with this；

Judge in the Candidate Set of valley point, whether the gray value of each extreme point meets the jitter elimination condition set one by one, if It is then this extreme point to be removed from the Candidate Set of valley point, otherwise, then retain this extreme point, finally give the candidate of non-jitter Collection, the described jitter elimination condition set as: | f (x_u)-f(x_u+1) | ＜ ε, wherein, xu and x_u+1It is respectively the Candidate Set of valley point In u extreme point and the u+1 extreme point, f (x_u) and f (x_u+1) it is respectively x_uAnd x_u+1Corresponding gray value, u=1,2 ..., p-1；P be valley point Candidate Set in the sum of extreme point, ε is the jitter parameter set；

Find out first from the Candidate Set of non-jitter and meet x_oFor maximum point and x_o+1Extreme point x for minimum point_o, so After with minimum point x_o+1Corresponding gray value f (x_o+1) as the rectangular histogram global threshold of source images, wherein, o=1,2 ..., q- 1；Q be non-jitter Candidate Set in the sum of extreme point, and q≤p.

Being further used as preferred embodiment, the described region to be filled according to robot uses path optimization's algorithm to enter Row region segmentation and optimum path planning, the step for of obtaining the optimal drawing path of robot, comprising:

Use the edge fitting method improved that the edge in the region to be filled of robot is fitted, obtain treating after matching Filling region, the edge fitting method employing multi-section circular arc of described improvement carrys out every section of curve at the edge in matching region to be filled, and Use the line segment obtained by connection two end points of this section of circular arc to approximate this when the radius of circular arc is less than the radius threshold set Section circular arc；

It is multiple inside subregion without cavity by the area to be filled regional partition after matching；

With the shortest total drafting time as target, use heuristic search algorithm determine each sub regions drawing order and The drafting starting point of each sub regions and drafting terminal, thus obtain the optimal drawing path of robot.

It is further used as preferred embodiment, described with the shortest total drafting time as target, use heuristic search to calculate Method determines the drawing order of each sub regions and the drafting starting point of each sub regions and draws terminal, thus obtains robot The step for of optimal drawing path, comprising:

Find the beeline between any two subregion；

Every sub regions is abstracted into a point, using the beeline between two sub regions as the weight on limit, will be the shortest Total drafting time is found problem and is converted into traveling salesman problem, then uses heuristic nearest neighbor algorithm to solve each sub regions Drawing order；

Calculating total drafting time in each pattern path in each sub regions, computing formula is:

$\left\{\begin{array}{c}{t}_{total}={\&Sigma;}_{r=1}^n{t}_r+(n-1)t_c\\ t_r=\left\{\begin{array}{c}2\sqrt{\frac{l_r}{a}},l_r<\frac{v^2}{a}\\ \frac{l_r}{v}+\frac{v}{a},l_r\&GreaterEqual;\frac{v^2}{a}\end{array}\right.\\ t_c=2\sqrt{\frac{width}{a}}\end{array}\right.,$

Wherein, t_totalFor always the drawing the time of present mode path in current sub-region, v be robot drawing instrument Big translational speed, a is the acceleration that robot drawing instrument is given, and n is present mode path base line in current sub-region Quantity, t_cMove to the time of next base line from a base line for robot drawing instrument_,t_rAnd l_rIt is respectively current son Filling time of r base line and length in present mode path in region, width is present mode road in current sub-region The width of footpath base line；

The shifting between different subregions is calculated with the translational speed of robot drawing instrument according to the distance between different subregions The dynamic time；

In drawing order according to each sub regions, each sub regions, the total of the path of each pattern draws time and difference Traveling time between subregion is modeled and optimization, obtains painting of each sub regions corresponding to the shortest total drafting time Starting point processed and drafting terminal, the shortest described total drafting time T_totalRelational expression be:

$\left\{\begin{array}{c}{T}_{total}=\underset{i,j}{min}\{D\&lsqb;n^{\&prime;},i\&rsqb;+T(cp\&lsqb;n^{\&prime;},i\&rsqb;,cp\&lsqb;n^{\&prime;},j\&rsqb;)\}\\ st\&lsqb;n^{\&prime;}\&rsqb;=cp\&lsqb;n^{\&prime;},i\&rsqb;\\ e\&lsqb;n^{\&prime;}\&rsqb;=cp\&lsqb;n^{\&prime;},j\&rsqb;\\ D\&lsqb;k,i\&rsqb;=\underset{j,t}{\min }\{D\&lsqb;k-1,i\&rsqb;+T(cp\&lsqb;k-1,j\&rsqb;,cp\&lsqb;k-1,t\&rsqb;)+T(cp\&lsqb;k-1,t\&rsqb;,cp\&lsqb;k,i\&rsqb;)\}\\ st\&lsqb;k-1\&rsqb;=cp\&lsqb;k-1,j\&rsqb;\\ e\&lsqb;k-1\&rsqb;=cp\&lsqb;k-1,t\&rsqb;\\ D\&lsqb;0,i\&rsqb;=0\end{array}\right.,$

Wherein, every sub regions is according to the corresponding group of drawing order, and n ' rolls into a ball for last, and k is kth group, k=1, 2 ..., n '；D [n ', i], D [k, i] and D [k-1, i] are respectively from the drafting starting point of first group a to n ', group k and group k-1 The shortest drafting time of some i, cp [n ', i] and cp [n ', j] is respectively group corner point i and j, st [n '] of n ' and e [n '] and divides Not Wei the drafting starting point of group n ' and draw terminal, T (cp [n ', i], cp [n ', j]) is the some j in the some i a to n ' in group n ' Drafting time, cp [k-1, j] and cp [k-1, t] are respectively the corner point i, T that corner point j and t, cp [k, i] is group k of group k-1 (cp [k-1, j], cp [k-1, t]) is the drafting time of the some t in the some j a to k-1 in group k-1, T (cp [k-1, t], cp [k, I]) it is the traveling time putting i in the some t a to k in group k-1, D [0, i] is the initial value of D [k, i].

Reference Fig. 1, the image drawing system of a kind of intelligent robot, including:

Input module, for inputting source images to be drawn；

Levy extraction module, for using the method processed based on mixing binary image that source images is carried out feature extraction, Obtain the region to be filled of robot；

Optimal drawing path acquisition module, uses path optimization's algorithm to carry out district for the region to be filled according to robot Regional partition and optimum path planning, obtain the optimal drawing path of robot；

Image rendering module, carries out image rendering for robot according to the optimal drawing path obtained.

Being further used as preferred embodiment, described optimal drawing path acquisition module includes:

Edge fitting unit, for using the edge fitting method of improvement to intend the edge in the region to be filled of robot Closing, obtain the region to be filled after matching, the edge fitting method of described improvement uses multi-section circular arc to come matching region to be filled Every section of curve at edge, and connection two end points gained of this section of circular arc are used when the radius of circular arc is less than the radius threshold set To line segment approximate this section of circular arc；

Cutting unit, being used for the area to be filled regional partition after matching is multiple inside subregion without cavity；

Optimal drawing path determines unit, for the shortest total drafting time as target, uses heuristic search algorithm true Determine the drawing order of each sub regions and the drafting starting point of each sub regions and draw terminal, thus obtaining the optimal of robot Drawing path.

Being further used as preferred embodiment, described optimal drawing path determines that unit includes:

Beeline finds subelement, for finding the beeline between any two subregion；

Drawing order solves subelement, for every sub regions being abstracted into a point, and the shortest with between two sub regions Searching of the shortest total drafting time problem as the weight on limit, is converted into traveling salesman problem by distance, then use heuristic recently Adjacent Algorithm for Solving goes out the drawing order of each sub regions；

Very first time computation subunit, total drafting time in each pattern path in calculating each sub regions, calculates Formula is:

$\left\{\begin{array}{c}{t}_{total}={\&Sigma;}_{r=1}^n{t}_r+(n-1)t_c\\ t_r=\left\{\begin{array}{c}2\sqrt{\frac{l_r}{a}},l_r<\frac{v^2}{a}\\ \frac{l_r}{v}+\frac{v}{a},l_r\&GreaterEqual;\frac{v^2}{a}\end{array}\right.\\ t_c=2\sqrt{\frac{width}{a}}\end{array}\right.,$

Wherein, t_totalFor always the drawing the time of present mode path in current sub-region, v be robot drawing instrument Big translational speed, a is the acceleration that robot drawing instrument is given, and n is present mode path base line in current sub-region Quantity, t_cMove to the time of next base line, t from a base line for robot drawing instrument_rAnd l_rIt is respectively current son Filling time of r base line and length in present mode path in region, width is present mode road in current sub-region The width of footpath base line；

Second Time Calculation subelement, for according to the distance between different subregions and the mobile speed of robot drawing instrument Degree calculates the traveling time between different subregion；

Draw starting point and drafting terminal determines unit, for according in the drawing order of each sub regions, each sub regions Traveling time between total drafting time in the path of each pattern and different subregion is modeled and optimization, obtains The drafting starting point of each sub regions that short total drafting time is corresponding and drafting terminal, the shortest described total drafting time T_totalRelation Expression formula is:

$\left\{\begin{array}{c}{T}_{total}=\underset{i,j}{min}\{D\&lsqb;n^{\&prime;},i\&rsqb;+T(cp\&lsqb;n^{\&prime;},i\&rsqb;,cp\&lsqb;n^{\&prime;},j\&rsqb;)\}\\ st\&lsqb;n^{\&prime;}\&rsqb;=cp\&lsqb;n^{\&prime;},i\&rsqb;\\ e\&lsqb;n^{\&prime;}\&rsqb;=cp\&lsqb;n^{\&prime;},j\&rsqb;\\ D\&lsqb;k,i\&rsqb;=\underset{j,t}{\min }\{D\&lsqb;k-1,i\&rsqb;+T(cp\&lsqb;k-1,j\&rsqb;,cp\&lsqb;k-1,t\&rsqb;)+T(cp\&lsqb;k-1,t\&rsqb;,cp\&lsqb;k,i\&rsqb;)\}\\ st\&lsqb;k-1\&rsqb;=cp\&lsqb;k-1,j\&rsqb;\\ e\&lsqb;k-1\&rsqb;=cp\&lsqb;k-1,t\&rsqb;\\ D\&lsqb;0,i\&rsqb;=0\end{array}\right.,$

Wherein, every sub regions is according to the corresponding group of drawing order, and n ' rolls into a ball for last, and k is kth group, k=1, 2 ..., n '；D [n ', i], D [k, i] and D [k-1, i] are respectively from the drafting starting point of first group a to n ', group k and group k-1 The shortest drafting time of some i, cp [n ', i] and cp [n ', j] is respectively group corner point i and j, st [n '] of n ' and e [n '] and divides Not Wei the drafting starting point of group n ' and draw terminal, T (cp [n ', i], cp [n ', j]) is the some j in the some i a to n ' in group n ' Drafting time, cp [k-1, j] and cp [k-1, t] are respectively the corner point i, T that corner point j and t, cp [k, i] is group k of group k-1 (cp [k-1, j], cp [k-1, t]) is the drafting time of the some t in the some j a to k-1 in group k-1, T (cp [k-1, t], cp [k, I]) it is the traveling time putting i in the some t a to k in group k-1, D [0, i] is the initial value of D [k, i].

Below in conjunction with Figure of description and specific embodiment the present invention it is further explained and illustrates.

Embodiment one

Reference Fig. 2-6, the first embodiment of the present invention:

Weak, less reliable and accurate for prior art robustness, draw the defect of time length, the present invention proposes one New robot drawing practice, carrys out the intelligence of show robot.For extracting characteristics of image, the present invention proposes a kind of based on the overall situation Binaryzation and the feature extracting method of local binarization technology mixing；Then use region segmentation method, complicated region is divided It is slit into simple subregion, the method finally combining a proposed by the invention optimum path planning, produces picture one width and draw Optimal path.

Fig. 2 is the general frame that phase system is drawn by robot proposed by the invention.Wherein, the image of input is to be clapped by camera The photochrome taken the photograph.But, the image drawn out is black white image, and the process ratio drawn is relatively time-consuming, so the figure drawn Succinct as should try one's best.Therefore, it is necessary to extract the feature that can give expression to source images.The feature that the present invention is extracted refers to one The closure region can filled by robot pen a bit.But, robot edges of regions after moving back and forth and filling It is probably jagged.Therefore, the edge in these regions smoothly must be drawn out by robot.Paint additionally, should optimize Figure path is to reduce the drafting time.The program of feature extraction and path planning is complex, and Gu Qihui transports on a personal computer OK.The path finally generated will be sent to robot, to perform drawing course.

Mainly feature extraction and path planning the two main process are described in detail below.

(1) feature extraction

The step that feature of present invention extracts, as it is shown on figure 3, first obtain the binary image of source images, is then converted into Vectogram.Binaryzation technology can be divided into two big classes: overall situation binaryzation--and use a fixing threshold value that whole image is carried out Binaryzation, and local binarization--use different threshold values to carry out binaryzation each pixel.In view of both approaches Cutting both ways, the present invention combines after the advantage of both approaches obtains binaryzation source images, then the bitmap-converted obtained For vectogram, to extract the geological information in each region.

(1) local binarization

Owing to the different piece of photo may have different illumination conditions, so the present invention needs to use local binarization skill Photo is processed by art.Proposing local auto-adaptive Binarization methods for this present invention, each pixel is calculated by this algorithm Going out a threshold value, specific algorithm process is as follows:

Step 1: calculate the arithmetic mean of instantaneous value of the gray scale of the peripheral region (i.e. neighborhood) of each pixel.

This process can be accelerated by using Integral Image (also referred to as Summed-area Table, integrogram) Realize to linear session.Assume in source images, point (x, gray value y) be p (x, y), then it is corresponding in integrogram Value I (x, y) is defined as follows:

$I(x,y)={\mathrm{\&Sigma;}}_{\underset{0\&le;j\&le;y}{0\&le;i\&le;x}}p(i,j)$

Whole Integral Image can be calculated within the linear time by below equation:

I (x, y)=p (x, y)+I (x-1, y)+I (x, y-1)-I (x-1, y-1)

The size assuming neighborhood is odd number s, and so (x, (x, y) can be by such as the following for the gray average m of neighborhood y) for point Formula calculates:

Sum (x, y)=I (x+d, y+d)-I (x+d, y-d-1)-I (x-d-1, y+d)+I (x-d-1, y-d-1)

$m(x,y)=sum(x,y)\&times;\frac{1}{s^2}$

Step 2: for each pixel calculate local threshold T (x, y), computing formula is:

T (x, y)=m (x, y)-c

Here c is a constant.

Step 3: the threshold value using Step 2 to calculate carries out binaryzation to image.The present invention is calculated by below equation Go out binary image each pixel b (x, y):

Local thresholding method performance in some aspects is as an edge detection method, therefore this method can be extracted well Picture shape feature.Obviously, size s of neighborhood and the resolution of source images are relevant.S is an odd number not less than 3.The present invention Propose an equation to calculate s:

Here, w and h is respectively width and the height of source images, α and β is given weight parameter

(2) overall situation binaryzation

Local binarization method in some aspects act like high pass filter, it can obtain very in some cases Strange result.Such as, one has large-area hair image robot may be made to draw one " bareheaded ".The present invention proposes Solution be by plus hair, repair " bareheaded " image, its step is as follows: 1) use local binary said before Change method, by source images binaryzation.2) overall situation binarization method is used, by source images binaryzation.3) this two images are merged.

The purpose of overall situation binaryzation is the dark portion region extracted and may contain hair.But, different images has different Illumination condition, and threshold value also should be otherwise varied, therefore the present invention proposes one determines algorithm based on histogrammic threshold value.Source figure One typical rectangular histogram of picture is as shown in Figure 4.In Fig. 4, some A is a preferable global threshold.

Present invention determine that the method for the threshold value of a histogram curve shaken is as follows:

Step 1: find all extreme points of histogram curve, and set up the Candidate Set of valley point with this.

Step 2: detect and eliminate shake.If x_uAnd x_u+1Meet inequality | f (x_u)-f(x_u+1) | ＜ ε, then scope So scope [x_u,x_u+1] part be one shake x_uAnd x_u+1Then need to remove from Candidate Set.Until examining the step for of repetition Till not detecting new shake.Here ε is the jitter parameter set,

Step 3: find out first from the Candidate Set of non-jitter and meet x_oFor maximum point and x_o+1Pole for minimum point Value point x_o, then with minimum point x_o+1Corresponding gray value f (x_o+1) as the rectangular histogram global threshold of source images.

(3) loss diagram

Binary picture after merging seems bitmap, and this figure is made up of some closure regions.But, in the bitmap of binaryzation Contained information only has the value (0 or 1) of each pixel.In order to obtain the information in each region, in addition it is also necessary to do some extra works Make.The more important thing is, robot needs to carry out retouching limit for each region, directly uses robot to point out each picture on border Plain and unwise.Therefore the path on border should be found, then allow robot move along this edge.

To this end, the scheme that the present invention proposes is to be vectogram bitmap-converted.Loss diagram by geometric expression formula rather than Pixel describes the vector outline in each region.After obtaining loss diagram, the present invention can be based on these with described by algebraic expression Border, come the path of planning robot.

The present invention uses Potrace software to complete the acquisition work of loss diagram.It addition, this software can also be by removing Too small region suppresses speckle, thus eliminates the noise in image and little details further.

(2) optimal path planning

Optimal path planning process can be further subdivided into:

(1) rim path matching

The edge that Potrace software produces is made up of some straight lines and Bezier curve.Bezier curve is a kind of at figure Widely used parameter curve in shape operation.But, most robot and CNC can only carry out straight line and circular motion.If When using straight line to remove matching Bezier curve, even if using substantial amounts of control point, it is also possible to cause the generation of edge sawtooth.This Invention proposes a kind of technology using multi-section circular arc to carry out one section of Bezier curve of matching, and the fitting result that it produces may be deposited At the circular arc of some minimum radiuses, and these circular arcs can not be processed by robot.In this case, the company of present invention uses Connect line segment obtained by two end points of circular arc to approximate this section of circular arc to avoid this problem, and will not lose too many accurate Degree.

(2) area filling

For filling the closure region after these matchings, the first step is by region segmentation.The shape in these closure regions Shape may be extremely complex, may comprise some cavities simultaneously.This can serve difficulty by band in ensuing path planning.This Bright by these region segmentation becoming simple, carries out simplification without empty subregion to solve this problem.

Then the path of every sub regions can be needed the drawing order determining every sub regions to obtain by independent generation To the shortest drafting time.For a region, fill path can start from any one corner or terminate, as shown in Figure 6. Starting point and end point can affect instrument traveling time when each multizone is filled.So, the present invention is firstly the need of determining this A little points.And for a region, it is possible to use different path modes, as shown in Figure 6.

(3) path planning

Path according to multizone connects theory, when connecting the path of zones of different, has two to affect the drafting time Key factor: 1) path mode of an intra-zone；2) displacement between zones of different.As it is shown in figure 5, we can be right The modeling of this problem is as follows: group g [k] is each region, and some cp [k, j] is the corner point of group k, some s [k-1] and e [k-1] minute It it not the beginning and end of a k-1.It is a NP-hard problem that the present invention finds the drawing path process of global optimum.This Bright employ a heuristic search to solve this problem, specifically comprise the following steps that

Step 1: find the beeline between all two sub regions.The step for can determine that theoretic interregional Beeline, in order to determine the priority drawing order of all subregion.Owing to corner point is considerably less, the step for can be by poor Act method is calculated.Actually also affected by drawing starting point and drafting terminal due to subregion, in addition it is also necessary to by Step 3 Obtain near-optimal solution and be used as the actual optimum solution of optimal path.

Step 2: by every sub regions as point, treat as limit with two the interregional beelines obtained in Step 1 Weight, find the shortest drawing order corresponding to the drafting time.This process is a traveling salesman problem, can pass through Heuristic nearest neighbor algorithm solves.

Step 3: determine the beginning and end of every sub regions.

As it was previously stated, the path mode of every sub regions and interregional displacement all can have influence on the drafting time.And The number of times that turns to of robot drawing instrument is the factor of the main instrument that an affects traveling time.And for the road of each pattern Footpath, can calculate its traveling time: the maximum translational speed assuming instrument is v, and acceleration is definite value a, scanning The quantity of row is n, and the length of each base line is l respectively₁,l₂,…l_n, the time filling each base line is t respectively₁,t₂,… t_n, the width of base line is width (this value typically the least), and the time moving to next base line from a base line is t_c, then total drafting time t_totalCan be calculated by below equation:

$\left\{\begin{array}{c}{t}_{total}={\&Sigma;}_{r=1}^n{t}_r+(n-1)t_c\\ t_r=\left\{\begin{array}{c}2\sqrt{\frac{l_r}{a}},l_r<\frac{v^2}{a}\\ \frac{l_r}{v}+\frac{v}{a},l_r\&GreaterEqual;\frac{v^2}{a}\end{array}\right.\\ t_c=2\sqrt{\frac{width}{a}}\end{array}\right.$

The drafting time in the path of each pattern can be calculated according to above formula, be relatively easy to because this is one Work.It addition, from the traveling time in a region to another one region, it is also possible to by similar calculating t_rMethod or according to Scanning speed and interregional distance are calculated.So, for every a pair beginning and end selected by every sub regions, all The time filling this region and the time shifted in the different areas can be calculated.

To this, the present invention is by as follows for optimization problem modeling: (i.e. the drafting of subregion is suitable for the traversal order of these groups given Sequence), some m from an i move to the point time T (cp [i, m], cp [j, n]) that spent of n in a j, it is thus necessary to determine that Mei Getuan Starting point st [k] of k and terminal e [k], to obtain the shortest overall travel time.Assume, from the starting point started most a to k Point i the shortest travel time be D (k, i).So n ', the shortest total travel time T are rolled into a ball for last_totalStarting point with it Can be calculated by below equation with terminal:

$T_{total}=\underset{i,j}{min}\{D\&lsqb;n^{\&prime;},i\&rsqb;+T(cp\&lsqb;n^{\&prime;},i\&rsqb;,cp\&lsqb;n^{\&prime;},j\&rsqb;)\}$

St [n ']=cp [n ', i]

E [n ']=cp [n ', j]

And D (k, value i) can pass through D (k-1, j) calculates, and computing formula is as follows:

$D\&lsqb;k,i\&rsqb;=\underset{j,t}{min}\{D\&lsqb;k-1,i\&rsqb;+T(cp\&lsqb;k-1,j\&rsqb;,cp\&lsqb;k-1,t\&rsqb;)+T(cp\&lsqb;k-1,t\&rsqb;,cp\&lsqb;k,i\&rsqb;)\}$

S [k-1]=v [k-1, j]

E [k-1]=cp [k-1, t]

The present invention by the D of first group (k, i) is set to 0, i.e. has:

D [0, i]=0

The time complexity more than calculated is 0 (k²N '), wherein, n ' is the quantity of group, and k is the corner point of each group Quantity.Because k is typically a least constant, so the time complexity of this path planning algorithm is low-down.

Embodiment two

Reference Fig. 7-10, the second embodiment of the present invention:

In order to verify the effect of the inventive method, the present invention proposes the experimental system shown in Fig. 7 to carry out experimental verification. As it is shown in fig. 7, this experimental system includes a camera, an ABB IRB120 industrial robot being equipped with marker.Mark Pen is connected in robot by a specially designed container, has a spring so that marker is on paper in this container Can be more smooth time mobile.Feature extraction and path planning algorithm then use C# to realize, and corresponding program is at a notes electricity Run on brain.

The detailed process of the present embodiment is as follows:

(1) feature extraction

The output of each step of feature extraction is as shown in the A-E of Fig. 8.Wherein, source images A is to use OpenCV program to enter Photo after row face extraction；B and C of Fig. 8 is local binarization and the result of overall situation binaryzation respectively, and the parameter of use is such as Under: a) c=9.0；B) α=β=0.04；C) ε=5, the empirical data that these experiments draw can make in most cases With.As shown in the B of Fig. 8, use the image after local binarization, can depict the shape of source images, but it obtains is one The strange result of individual little hair.Therefore, also need to repair this result by increase hair.As shown in the C of Fig. 8, pass through Overall situation binaryzation, can extract the dark portion region comprising hair.Then, this two pictures of B and C is merged by we, with Obtaining last binaryzation result (i.e. the D of Fig. 8), the bitmap D after then using Potrace software to merge is converted to vectogram E.During conversion, the parameter of Potrace software is "-t 10-s ", and this represents that turdsize when suppressing speckle is 10, output format For SVG.The E of Fig. 8 is the vectogram after conversion, and this is also the output that feature extraction is final.

(2) path planning

The F of Fig. 8 is the result that edge carries out circular fitting, it comprises 845 sections of straight lines or circular arc.It can be seen that intend Result and original vectogram after conjunction are the most identical.

After circular fitting, these closure regions can be divided into subregion.After obtaining suitable subregion, can first use The path join algorithm of the present invention obtains the drawing order of these subregions, determines starting point and the end of every sub regions the most again Point, to obtain the shortest drafting time.Finally, generate whole drawing path and path is sent to robot.Fig. 9 illustrates machine The drawing course of device people, wherein, A is by edge circular fitting, and B is by area filling.

After using the method for the present invention, the final image that robot use marker draws is as shown in Figure 10.Permissible Finding out, the effect of the image that robot draws automatically is the prettyst good.As can be seen here, feature extraction proposed by the invention and road The effect of footpath planning algorithm is preferable.Meanwhile, the method for the present invention is likely to be applied to other commercial Application and (such as sprays paint and polish Deng) in, to control robot intelligently to complete required task.

The present invention proposes an intelligent robot using image characteristics extraction and path planning algorithm to carry out drawing System and its implementation.For processing the photo with different illumination conditions, present invention uses based on local binarization with complete The mixed method of office's binaryzation carries out feature (i.e. closure region) and extracts；Because robot uses to move back and forth fills one piece During region, jagged edge can be caused, so the present invention uses the information of vectogram to first pass through circular fitting by these regions Carry out retouching limit to obtain the edge smoothed；For obtaining the shortest drafting time, the present invention proposes a new path optimization and calculates Method, the path of this algorithm coordinates measurement based on single region and multizone connects theory.Test result indicate that, side of the present invention Method can carry out image rendering effectively.Feature extraction proposed by the invention and the method for optimum path planning, not merely can answer It can also be used to many needs in the commercial Application of intelligent robot technology in robot draws, as sprayed paint and polishing etc..

It is above the preferably enforcement of the present invention is illustrated, but the invention is not limited to described enforcement Example, those of ordinary skill in the art also can make all equivalent variations on the premise of spirit of the present invention or replace Changing, deformation or the replacement of these equivalents are all contained in the application claim limited range.

Claims (10)

1. the image drawing method of an intelligent robot, it is characterised in that: comprise the following steps:

Input source images to be drawn；

Use the method processed based on mixing binary image that source images is carried out feature extraction, obtain the area to be filled of robot Territory；

Region to be filled according to robot uses path optimization's algorithm to carry out region segmentation and optimum path planning, obtains machine The optimal drawing path of people；

Robot carries out image rendering according to the optimal drawing path obtained.

The image drawing method of a kind of intelligent robot the most according to claim 1, it is characterised in that: described employing based on The method that mixing binary image processes carries out feature extraction to source images, obtains bag the step for of the region to be filled of robot Include:

Use local binarization method that source images is processed, obtain local binarization image；

Use overall binaryzation method based on histogram thresholding that source images is processed, obtain overall situation binary image；

Local binarization image and overall situation binary image are merged, obtains including the two of the region to be filled of robot Value image；

Binary image after merging is vectogram by bitmap-converted.

The image drawing method of a kind of intelligent robot the most according to claim 2, it is characterised in that: described employing local Source images is processed by binaryzation method, the step for of obtaining local binarization image, comprising:

Using integrogram method to calculate the gray scale arithmetic mean of instantaneous value of the neighborhood of each pixel in source images, computing formula is:

Wherein, p (x, y) be source images midpoint (x, pixel value y), I (x, y) be source images midpoint (x, y) right in integrogram The pixel value answered, s be source images midpoint (x, the size of neighborhood y),For rounding downwards symbol, w and h is respectively source The width of image and height, α and β is given weight parameter, and (x is y) that (x, the gray scale of neighborhood y) is calculated at source images midpoint to m Art meansigma methods；

Each pixel that meansigma methods is source images according to calculating calculates corresponding local threshold, and computing formula is:

((x, y)-c, wherein, (x is y) that (c is given constant for x, local threshold y) at source images midpoint to T to T for x, y)=m；

Local threshold according to calculating carries out local threshold to source images, obtains local binarization image, described source images Local threshold formula is:

Wherein, (x y) is source images midpoint (x, y) pixel value after local threshold to b.

The image drawing method of a kind of intelligent robot the most according to claim 2, it is characterised in that: described employing based on Source images is processed by the overall binaryzation method of histogram thresholding, the step for of obtaining overall situation binary image, comprising:

Obtain the histogram curve of source images；

Histogram curve according to source images uses jitter elimination method to determine the rectangular histogram global threshold of source images；

According to the global threshold determined, source images is carried out binaryzation, obtain overall situation binary image.

The image drawing method of a kind of intelligent robot the most according to claim 4, it is characterised in that: described according to source figure The histogram curve employing shake diagnostic method of picture includes the step for of determining the rectangular histogram global threshold of source images:

Find all extreme points of the histogram curve of source images, and set up the Candidate Set of valley point with this；

Judge in the Candidate Set of valley point, whether the gray value of each extreme point meets the jitter elimination condition set, if so, one by one Then this extreme point is removed from the Candidate Set of valley point, otherwise, then retain this extreme point, finally give the Candidate Set of non-jitter, The described jitter elimination condition set as: | f (x_u)-f(x_u+1) | ＜ ε, wherein, x_uAnd x_u+1It is respectively u in the Candidate Set of valley point Individual extreme point and the u+1 extreme point, f (x_u) and f (x_u+1) it is respectively x_uAnd x_u+1Corresponding gray value, u=1,2 ..., p-1；p For the sum of extreme point in the Candidate Set of valley point, ε is the jitter parameter set；

Find out first from the Candidate Set of non-jitter and meet x_oFor maximum point and x_o+1Extreme point x for minimum point_o, then with Minimum point x_o+1Corresponding gray value f (x_o+1) as the rectangular histogram global threshold of source images, wherein, o=1,2 ..., q-1；q For the sum of extreme point in the Candidate Set of non-jitter, and q≤p.

6. according to the image drawing method of a kind of intelligent robot described in any one of claim 1-5, it is characterised in that: described Region to be filled according to robot uses path optimization's algorithm to carry out region segmentation and optimum path planning, obtains robot The step for of optimal drawing path, comprising:

Using the edge fitting method improved to be fitted the edge in the region to be filled of robot, obtain after matching is to be filled Region, the edge fitting method employing multi-section circular arc of described improvement carrys out every section of curve at the edge in matching region to be filled, and is justifying The line segment obtained by connection two end points of this section of circular arc is used to approximate this Duan Yuan when the radius of arc is less than the radius threshold set Arc；

It is multiple inside subregion without cavity by the area to be filled regional partition after matching；

With the shortest total drafting time as target, use heuristic search algorithm determine the drawing order of each sub regions and each The drafting starting point of subregion and drafting terminal, thus obtain the optimal drawing path of robot.

The image drawing method of a kind of intelligent robot the most according to claim 6, it is characterised in that: described with the shortest always The drafting time is target, uses heuristic search algorithm to determine drawing order and the drafting of each sub regions of each sub regions Starting point and drafting terminal, thus the step for of obtaining the optimal drawing path of robot, comprising:

Find the beeline between any two subregion；

Every sub regions is abstracted into a point, using the beeline between two sub regions as the weight on limit, will the shortest always paint Time processed is found problem and is converted into traveling salesman problem, then uses heuristic nearest neighbor algorithm to solve the drafting of each sub regions Sequentially；

Calculating total drafting time in each pattern path in each sub regions, computing formula is:

$\left\{\begin{array}{c}{t}_{total}={\mathrm{\&Sigma;}}_{r=1}^n{t}_r+(n-1)t_c\\ t_r=\left\{\begin{array}{c}2\sqrt{\frac{l_r}{a}},l_r<\frac{v^2}{a}\\ \frac{l_r}{v}+\frac{v}{a},l_r\&GreaterEqual;\frac{v^2}{a}\end{array}\right.\\ t_c=2\sqrt{\frac{width}{a}}\end{array}\right.,$

Wherein, t_totalFor always the drawing the time of present mode path in current sub-region, v is the maximum shifting of robot drawing instrument Dynamic speed, a is the acceleration that robot drawing instrument is given, and n is the quantity of present mode path base line in current sub-region, t_cMove to the time of next base line, t from a base line for robot drawing instrument_rAnd l_rIt is respectively current sub-region Filling time of r base line and length in interior present mode path, width is that present mode path is swept in current sub-region Retouch capable width；

According to the distance between different subregions calculate between different subregions with the translational speed of robot drawing instrument mobile time Between；

In drawing order according to each sub regions, each sub regions, the total of the path of each pattern draws time and the most sub-district Traveling time between territory is modeled and optimization, and the drafting obtaining each sub regions corresponding to the shortest total drafting time rises Point and drafting terminal, the shortest described total drafting time T_totalRelational expression be:

$\left\{\begin{array}{c}{T}_{total}=\underset{i,j}{min}\{D\&lsqb;n^{\&prime;},i\&rsqb;+T(cp\&lsqb;n^{\&prime;},i\&rsqb;,cp\&lsqb;n^{\&prime;},j\&rsqb;)\}\\ st\&lsqb;n^{\&prime;}\&rsqb;=cp\&lsqb;n^{\&prime;},i\&rsqb;\\ e\&lsqb;n^{\&prime;}\&rsqb;=cp\&lsqb;n^{\&prime;},j\&rsqb;\\ D\&lsqb;k,i\&rsqb;=\underset{i,j}{min}\{D\&lsqb;k-1,i\&rsqb;+T(cp\&lsqb;k-1,j\&rsqb;,cp\&lsqb;k-1,t\&rsqb;)+T(cp\&lsqb;k-1,t\&rsqb;,cp\&lsqb;k,i\&rsqb;)\}\\ st\&lsqb;k-1\&rsqb;=cp\&lsqb;k-1,j\&rsqb;\\ e\&lsqb;k-1\&rsqb;=cp\&lsqb;k-1,j\&rsqb;\\ D\&lsqb;0,i\&rsqb;=0\end{array}\right.,$

Wherein, every sub regions is according to the corresponding group of drawing order, and n ' rolls into a ball for last, and k is kth group, k=1, 2 ..., n '；D [n ', i], D [k, i] and D [k-1, i] are respectively from the drafting starting point of first group a to n ', group k and group k-1 The shortest drafting time of some i, cp [n ', i] and cp [n ', j] is respectively group corner point i and j, st [n '] of n ' and e [n '] and divides Not Wei the drafting starting point of group n ' and draw terminal, T (cp [n ', i], cp [n ', j]) is the some j in the some i a to n ' in group n ' Drafting time, cp [k-1, j] and cp [k-1, t] are respectively the corner point i, T that corner point j and t, cp [k, i] is group k of group k-1 (cp [k-1, j], cp [k-1, t]) is the drafting time of the some t in the some j a to k-1 in group k-1, T (cp [k-1, t], cp [k, I]) it is the traveling time putting i in the some t a to k in group k-1, D [0, i] is the initial value of D [k, i].

8. the image drawing system of an intelligent robot, it is characterised in that: including:

Input module, for inputting source images to be drawn；

Levy extraction module, for using the method processed based on mixing binary image that source images is carried out feature extraction, obtain The region to be filled of robot；

Optimal drawing path acquisition module, uses path optimization's algorithm to carry out region for the region to be filled according to robot and divides Cut and optimum path planning, obtain the optimal drawing path of robot；

Image rendering module, carries out image rendering for robot according to the optimal drawing path obtained.

The image drawing system of a kind of intelligent robot the most according to claim 8, it is characterised in that: described optimal drafting Path acquisition module includes:

Edge fitting unit, for using the edge fitting method of improvement that the edge in the region to be filled of robot is fitted, Obtaining the region to be filled after matching, the edge fitting method of described improvement uses multi-section circular arc to come the edge in matching region to be filled Every section of curve, and use when the radius of circular arc is less than the radius threshold set and connect obtained by two end points of this section of circular arc Line segment approximates this section of circular arc；

Cutting unit, being used for the area to be filled regional partition after matching is multiple inside subregion without cavity；

Optimal drawing path determines unit, for the shortest total drafting time as target, uses heuristic search algorithm to determine respectively The drawing order of sub regions and the drafting starting point of each sub regions and draw terminal, thus obtain the optimal drafting of robot Path.

The image drawing system of a kind of intelligent robot the most according to claim 9, it is characterised in that: described most preferably paint Path determining unit processed includes:

Beeline finds subelement, for finding the beeline between any two subregion；

Drawing order solves subelement, for every sub regions being abstracted into a point, with the beeline between two sub regions As the weight on limit, searching of the shortest total drafting time problem is converted into traveling salesman problem, then uses heuristic arest neighbors to calculate Method solves the drawing order of each sub regions；

Very first time computation subunit, total drafting time in each pattern path, computing formula in calculating each sub regions For:

$\left\{\begin{array}{c}{t}_{total}={\mathrm{\&Sigma;}}_{r=1}^n{t}_r+(n-1)t_c\\ t_r=\left\{\begin{array}{c}2\sqrt{\frac{l_r}{a}},l_r<\frac{v^2}{a}\\ \frac{l_r}{v}+\frac{v}{a},l_r\&GreaterEqual;\frac{v^2}{a}\end{array}\right.\\ t_c=2\sqrt{\frac{width}{a}}\end{array}\right.,$

Wherein, t_totalFor always the drawing the time of present mode path in current sub-region, v is the maximum shifting of robot drawing instrument Dynamic speed, a is the acceleration that robot drawing instrument is given, and n is the quantity of present mode path base line in current sub-region, t_cMove to the time of next base line, t from a base line for robot drawing instrument_rAnd l_rIt is respectively current sub-region Filling time of r base line and length in interior present mode path, width is that present mode path is swept in current sub-region Retouch capable width；

Second Time Calculation subelement, based on the translational speed according to the distance between different subregions and robot drawing instrument Calculate the traveling time between different subregion；

Draw starting point and drafting terminal determines unit, for according in the drawing order of each sub regions, each sub regions every kind Traveling time between total drafting time in the path of pattern and different subregion is modeled and optimization, obtain the shortest always The drafting starting point of each sub regions that the drafting time is corresponding and drafting terminal, the shortest described total drafting time T_totalRelationship expression Formula is:

$\left\{\begin{array}{c}{T}_{total}=\underset{i,j}{min}\{D\&lsqb;n^{\&prime;},i\&rsqb;+T(cp\&lsqb;n^{\&prime;},i\&rsqb;,cp\&lsqb;n^{\&prime;},j\&rsqb;)\}\\ st\&lsqb;n^{\&prime;}\&rsqb;=cp\&lsqb;n^{\&prime;},i\&rsqb;\\ e\&lsqb;n^{\&prime;}\&rsqb;=cp\&lsqb;n^{\&prime;},j\&rsqb;\\ D\&lsqb;k,i\&rsqb;=\underset{i,j}{min}\{D\&lsqb;k-1,i\&rsqb;+T(cp\&lsqb;k-1,j\&rsqb;,cp\&lsqb;k-1,t\&rsqb;)+T(cp\&lsqb;k-1,t\&rsqb;,cp\&lsqb;k,i\&rsqb;)\}\\ st\&lsqb;k-1\&rsqb;=cp\&lsqb;k-1,j\&rsqb;\\ e\&lsqb;k-1\&rsqb;=cp\&lsqb;k-1,j\&rsqb;\\ D\&lsqb;0,i\&rsqb;=0\end{array}\right.,$

Wherein, every sub regions is according to the corresponding group of drawing order, and n ' rolls into a ball for last, and k is kth group, k=1, 2 ..., n '；D [n ', i], D [k, i] and D [k-1, i] are respectively from the drafting starting point of first group a to n ', group k and group k-1 The shortest drafting time of some i, cp [n ', i] and cp [n ', j] is respectively group corner point i and j, st [n '] of n ' and e [n '] and divides Not Wei the drafting starting point of group n ' and draw terminal, T (cp [n ', i], cp [n ', j]) is the some j in the some i a to n ' in group n ' Drafting time, cp [k-1, j] and cp [k-1, t] are respectively the corner point i, T that corner point j and t, cp [k, i] is group k of group k-1 (cp [k-1, j], cp [k-1, t]) is the drafting time of the some t in the some j a to k-1 in group k-1, T (cp [k-1, t], cp [k, I]) it is the traveling time putting i in the some t a to k in group k-1, D [0, i] is the initial value of D [k, i].